(111a) A Data-Driven Approach for Modeling and Energy Optimization of Smr Hydrogen Production Process

Recently, the attention of hydrogen as a clean energy source has increased, because it can deal with various issues related to the fossil fuel. The hydrogen is usually supplied by three primary methods and the steam methane reforming(SMR) process reforming the natural gas is the most common method. In the case of small scale of SMR process, the SMR reactor has a complex structure such as double-tube type of reactors. The complexity has limitations in expressing the systems with several assumptions. This study focuses on modeling the SMR process by data-driven approach and optimization for maximizing the thermal efficiency of the process using intelligent optimization algorithms. A deep neural network(DNN) has been constructed to find the relations between the process variables and trained by actual operation data. To improve the model accuracy, the noise and outliers in the data are removed by data preprocessing methods. The optimization of neural networks which are black box models is implemented in a heuristic way to find the optimal point of the system. However, the result highly depends on the type of optimization algorithm and it is important to select the most appropriate one for the system. Three optimization algorithms are used to search for the optimal solution: grid search(GS), genetic algorithm(GA), and particle swarm optimization(PSO). As a result, the GS shows the fastest speed to find the optimal solution among them because the it has no iterative calculations. Unfortunately, when the number of decision variables increases, the calculation cost has greatly increased and optimization is not available. On the other hand, the number of decision variables is not limited in the GA and PSO. The thermal efficiency of the process increases as the number of decision variable increases, but the search time increases significantly in GA. Therefore, the PSO shows the best performance to optimize the SMR process. Consequently, the optimal thermal efficiency of 85.4% is obtained and this result will contribute to increase the productivity of hydrogen.